Q

QVQ 72B Preview AWQ

Developed by kosbu
QVQ-72B-Preview is an experimental research model developed by the Qwen team, focusing on enhancing visual reasoning capabilities. This repository provides its AWQ 4-bit quantized version.
Downloads 532
Release Time : 12/24/2024

Model Overview

This model is a visual reasoning enhancement model developed by the Qwen team, optimized with AWQ 4-bit quantization technology, supporting multi-GPU tensor parallelism. It demonstrates excellent performance in multiple visual reasoning benchmarks.

Model Features

Efficient Quantization
Utilizes AWQ 4-bit quantization technology, significantly reducing model size and computational resource requirements
Multi-GPU Compatibility
Addresses divisibility constraints through zero-padding, enabling efficient tensor parallelism across multiple GPUs
Enhanced Visual Reasoning
Demonstrates outstanding performance on visual reasoning benchmarks such as MMMU and MathVista

Model Capabilities

Visual Reasoning
Cross-modal Understanding
Mathematical Problem Solving
Multidisciplinary Knowledge Application

Use Cases

Education
Mathematical Problem Solving
Analyzing questions containing mathematical formulas and diagrams
Achieves 71.4% accuracy on the MathVista benchmark
Research
Multimodal Research
Used for research in visual language understanding and reasoning tasks
Achieves 70.3% accuracy on the MMMU benchmark
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase